Cancer clinical trials are often subject to misclassification of discrete covariates, measurement errors in continuous covariates, categorization of covariates, model misspecification, and omission of prognositc factors in the analysis. Each of these "errors" increase the variance of the estimate of the treatment effect thereby increasing the required sample size or increasing the level of significance. The aims of the proposed study are to quantify the effect these "errors" have on the precision of the estimate for treatment effect, adjusted for covariates under Cox's proportional hazards model.